High Efficiency Video Coding (HEVC) is a video compression standard currently under development by the Joint Collaborative Team on Video Coding (JCT-VC) formed by ISO/IEC Moving Picture Experts Group (MPEG) and ITU-T Video Coding Experts Group (VCEG). In the present HEVC specification, a picture is coded in a Largest Coding Unit (LCU). An LCU may be a 128×128 block, a 64×64 block, a 32×32 block or a 16×16 block. An LCU can be encoded directly or may be divided into four Coding Units (CUs) for next level encoding. A CU may be encoded directly or may be further divided into a next level for encoding. The smallest CU is an 8×8 block.
In general, at each level, a CU of size 2N×2N may be divided into Prediction Units (PU) for the purposes of prediction. For intra coding, a 2N×2N CU may be encoded in one 2N×2N PU or into four N×N PUs. For inter coding, a 2N×2N CU may be encoded in one 2N×2N PU, two 2N×N PUs, two N×2N PUs, a 0.5N×2N PU plus a 1.5N×2N PU, a 1.5N×2N PU plus a 0.5N×2N PU, a 2N×0.5N PU plus a 2N×1.5N PU, a 2N×1.5N PU plus a 2N×0.5N PU, or four N×N PUs. In color video where picture data may be distributed over three channels including a luminance (luma) channel Y and two chrominance (chroma) channels U and V, a PU may include one luma block Y and two chroma blocks U and V.
In an HEVC encoder, after intra prediction (Intra-frame Prediction module) or inter prediction (Motion Estimation & Motion Compensation modules) are performed, the prediction residuals corresponding to the difference between an input PU and the predicted PU are transformed and quantized for entropy coding. When a PU is encoded in intra coding mode, different intra prediction modes may be applied including DC prediction, planar prediction, horizontal prediction, vertical prediction and so forth.